Introduction

This page gives a small introduction to MAL. First, we give a brief explanation of how MAL fits into a projection, then we explain how model data in MAL, and at last we examine an example program.

Projection

Below you see a flowchart modeling the projection of a portfolio. The flowchart considers three types of portfolio data, namely:

• Input coming from a database (during testing from some Excel sheets).
• State produced by an init and manage step.
• Result produced by a projection step.

The purpose of MAL is to initialize a portfolio during the initialize portfolio state* step and subsequently managing it during the mange steps.

The initialize portfolio state step uses the input to initialize a portfolio state. The state both contains values required by the projection and custom values company specific values. As long as the projection is incomplete, PROBABLI will perform projection step which produces a projection result from a state. This result is used during manage portfolio state to update the portfolio state. When the projection is done all value traces are collected during finalize output and reported in an Excel sheet.

Projection data exists on the following six levels: Equity, Group, Policy, CashFlow, Asset, and Global. Each level contains an data entity modeling the data on the level. On some levels there exists several entities, e.g. there exist three kinds Group each modeling a different type of company grouping. All level entities have data on the following four levels:

• Input is the data coming from the database which is only available during initialize portfolio state.
• State is the data directly modifiable and is specified by MAL. Some of the state data is used by the projection and must therefore exists, e.g. a Policy must have a Reserve
• Result is the data coming from a projection step which is only available during manage portfolio state.
• Param is execution time parameters provided by the user.

Data modeling

The first part of a MAL program models the data used in the projection using a data definition. Quite a bit of data-modeling is needed since the projection uses a lot of data, see demo.fmal for a minimal model. A Group could be modeled as follows.

data Group
        Input     : GroupInput
        Result    : GroupResult
        Param     : GroupParam
        Assets    : List<Asset>
        Policies  : List<Policy>
end

Which states that a group contains five fields, with the type of the fields to the right of the :. If we want to model an Interest group we do so by extending the Group definition. This means that all Interest has all the fields of a Group plus some.

data InterestParam extends GroupParam
        InterestRate : Float
end

data Interest extends Group
        Param          : InterestParam
        InterestFields : Float, Output as CashFlow(Debug)
end

We are allowed to extend the fields of an extended entity. For example Interest use an extended field Param of the type InterestParam. In this case InterestParam must extend GroupParam.

Programs

An MAL-program consists of a data-declarations, function declarations, an init block and an update block. A small program is seen below:

/* Data definitions */

init
  update group in Groups
  with
    group.Reserve = group.Input.Reserve
  end
end

manage
  update group in Groups
  with
    group.Reserve = group.Result.Reserve
  end
end

Say want to initialize the a groups reserve differently depending on which kind of group, then we can use a match expression.

init
  update group in Groups
  with
    group.Reserve =
      match group with
      | Interest iGrp -> 1
      | Expense EGrp -> 2
      | Risk rGrp -> 3
      end
  end
end

We could also only want to manage interest or expense groups.

manage
  update group in Groups:{Interest, Expense}
  with
    group.Reserve =
      match group with
      | Interest iGrp -> 1
      | Expense EGrp -> 2
      end
  end
end

Or we could want the group reserve to be the sum of its policies reserve after management.

manage
  update group in Groups
  with
    let polReserves = map p in group.Policies with p.Reserve end
    group.Reserve = sum(polReserve)
  end
end

Or only the positive reserves

manage
  update group in Groups
  with
    let polReserves = map p in group.Policies where p.reserve > 0 with p.Reserve end
    group.Reserve = sum(polReserve)
  end
end